In the realm of data clustering, the Deep Embedded Clustering (DEC) algorithm has earned a reputation for efficiently grouping data points. Its limitation is that it only deals with numerical data. In real-world scenarios, data is often a mixture of numerical and categorical attributes, posing a more intricate challenge. This project presents an enhanced version of the DEC framework, tailored to address the complexities of mixed data clustering. It incorporates embedded layers and soft-target updates to ensure seamless handling of both numerical and categorical attributes, maintaining convergence stability throughout the process. It also uses the concept of a “deep reinforcement learning” In the evaluation process, the proposed approach performed better than standard metrics.